Another example of the wonders of the internet: the Metathesis website, whose centerpiece is the Metathesis Database [run by Beth Hume].

What is metathesis? Metathesis is the phenomenon whereby two sounds that appear in a particular order in one form of a word occur in the reverse order in a related form of the word. […] The goal of this research project is two-fold. The first is to provide a more solid empirical basis for the study of metathesis. To achieve this, we are developing a database of reported cases of metathesis. […] (Note that not all reported cases of metathesis are actual cases of metathesis, as noted in some of the language listings.) The second aim of this project is to come to a clearer understanding of the nature of metathesis and, with this knowledge, develop a constrained and predictive theory of metathesis.

What a treasure for linguists! (Hat tip to Paul for the link.)

Update (Sept. 2023). The last snapshot preserved by the Internet Archive is Sept. 4, 2014. Sic tarnsit…


  1. Seems fairly ancient, ne c’est pas? It’s from 2000, and yet barely half the languages cataloged have pages or been updated.

  2. From the description, the first example that came to mind in North American English was the use of “ax” for “ask”, which is common in AAVE.
    The website doesn’t have Irish, so I couldn’t look for a similar example, which is the word for “box” which could be either “bocsa” or “bosca”.
    Am I missing something and these are not true examples of metathesis? And am I missing something else, because the list of examples seemed pretty short?

  3. If Dumbarton means Fort of the British, why isn’t it Dumbraton?

  4. If you have to pronounce Dumbraton 100 times every day, I’d dare say it’ll get shortened to Dumbarton very soon – in twenty centuries or so….

  5. Seems fairly ancient, ne c’est pas? It’s from 2000, and yet barely half the languages cataloged have pages or been updated.
    Hell, I didn’t notice that. Well, consider this an obituary post, then.

  6. Garrigus Carraig says

    I think ax comes from OE acsian, axian, which was a variant of ascian. You can see it in the examples under ascian here. The Metathesis database says of OE, “Medial sk often underwent metathesis to ks in late West Saxon.”

  7. maidhc, Garrigus: Indeed, the metathesis of ask is very old in English, though it’s unclear whether AAVE ax is a survival or a re-creation. However, the cognate words make it clear that ask is the original form.

  8. You may enjoy some contrepèteries in English (contrepets are demanding form of meaningfula nd purposeful spoonerisms)

  9. Dear Governor Brown [Jerry Brown’s father]:

    Can you please change the name of the Dumbarton Bridge?


    Mary Barton

  10. Here’s a fine (authentic) spoonerism: “skerched orth”.

  11. Sic tarnsit…

    Not only the database, but the languages. Rotuman, of which it has been observed “Word stress is associated with left-dominant bimoraic feet.” is endangered.

  12. David Eddyshaw says

    Although what is going on in Rotuman is often described as “metathesis”, I think it’s really a two-step process of umlaut followed by final vowel loss, though I suppose it all comes to the same thing in the end.

    It’s of interest to me as being remarkably similar to the behaviour of Kusaal “prosodic enclitics”, including the fact that what looks like a purely phonological process constantly behaves like an enclitic particle. (There is a relevant parallel in its fairly close relation Tongan, where shift of stress to the word-final syllable works like a definite article.)

    As an example of what I mean from Kusaal:

    Li ka’ daugɔ. “It’s not a log.”
    Li anɛ daug. “It’s a log.”

    Here daugɔ is actually realised /da:’gɔ/, whereas daug is /daʊg/. The proto-Western-Oti-Volta form was *da:gʊ.

    The longer form of Rotuman words (Churchward’s “Complete Phase”) behaves syntactically as if it were followed by a “definite article”, like the final-stressed forms in Tongan. I suspect that, historically, the Complete Phase had a final stress which preserved its original final vowel, whereas the Incomplete Phase had penult stress. Something similar probably underlies the alternation in Kusaal, though the now-disappeared enclitic responsible for the stress shift was a negative particle rather than an article.

    I actually toyed with adopting Churchward’s terms “Complete” and “Incomplete Phase” for what I just call Kusaal “Long” and “Short Forms” in my grammar, before reluctantly deciding that Rotuman might be somewhat obscure as a reference point. (I have a copy of his Rotuman Grammar and Dictionary, which is really quite wonderful.)

    Rotuman is also interesting in that it has a stratified vocabulary: there are huge numbers of borrowings from Polynesian, recognisable because they have not undergone some of the extremely distinctive historical sound changes seen in echt Rotuman.

  13. David Marjanović says

    See also: nominalization of adjectives by stress shift to the first syllable in Indo-European. (That’s the main mechanism for getting stressed syllabic consonants/zero-grades.)

  14. David Eddyshaw says

    Interesting to see Kasem on the list, as I know a bit about that one. Unfortunately, the ascription of metathesis to Kasem is just Chomsky and Halle bollocks of the kind that whole book of theirs so thoroughly exemplifies.

  15. David Marjanović says

    Here it is, and it seems to consist of positing it as an intermediate step for why apparently expected |°i-a| comes out as /e/. If that’s all, it’s entirely theoretical at best. Direct *-[ja]- > -[ɛː]- is known elsewhere, e.g. in OHG where positing an intermediate metathesis would greatly complicate things.

  16. David Eddyshaw says

    Mooré actually has a real live visible-to-the-naked-eye negative enclitic particle wherever Kusaal has its “prosodic enclitic” negative particle:

    Yaa raoogo.
    “It’s a log.”

    Ka raoog ye.
    “It’s not a log.”

    (Mooré also deletes underlying final short vowels, but it just does so whenever a word is not clause-final or prepausal, quite unlike Kusaal with its cross-linguistically weird Rotumanoid interactions with the syntax.)

  17. David Eddyshaw says

    Here it is

    Interestingly, the Kasem “sheep” word actually has a perfectly good Central Gur etymology, which shows unequivocally that the stem is in fact *pe-: cf Kusaal pɛ’og, Yom pɛ́ɣʊ̀, Nawdm fɛ́ɛ́gú. From the more closely related also-Grusi language Kabiyè, there is the less obvious but still cognate heu “sheep” (Kabiyè h- for *p- is regular.)

    The relevant sound change is actually the cross-linguistically unremarkable *pea -> pia in the singular. No metathesis here, nothing to see, move along please …

    [The singular for “sheep” in Kasem is actually piə; the plural is usually peeni, though pe does occur.]

    For Chomsky, data exists only to be tormented until it fits the theory. A perfect pseudoscientist.

  18. Although what is going on in Rotuman is often described as “metathesis”, I think it’s really a two-step process of umlaut followed by final vowel loss, though I suppose it all comes to the same thing in the end.

    Blevins and Garrett, The origins of consonant-vowel metathesis (here), §3 (Compensatory metathesis) and especially §3.2, look at Rotuman in detail. It’s kind of what you say, except there is not always umlaut. Sometimes you just end up with a sequence of two vowels.

  19. David Eddyshaw says

    Very interesting paper: thanks, Y.

    “Umlaut” is a bit of an oversimplfication for what I actually had in mind: vowel lengthening in a stressed open syllable, with the second mora of the lengthened vowel harmonising with the following vowel across the intervening consonant.
    Then the unstressed final vowel was deleted.

    All of that can actually be paralleled in Kusaal (for example, long vowels in the closed syllables resulting from final vowel deletion are “longer” in morphophonemic terms than long vowels in open syllables, so that CV:C syllables can carry two tones and never lose stress, whereas CV: syllables only carry one tone and lose their stress before a following stressed syllable.)

    The main difference in Kusaal is that the “harmonising” part only takes place across velar consonants and for the most part is confined to rounding assimilation (though fronting also probably crossed *ɣ before it was later lost completely.)

    But, in line with my generally troglodyte-nominalist theoretical position, I can’t really get worked up over whether one calls the result of all this “metathesis” or not. I do feel in my bones that absolutely regular metathesis calls for some explaining, though, if not actual explaining away as being “really” something a bit less outlandish.

Speak Your Mind